Thesis
Reference
Essays on polarization, multidimensional poverty and health-related effects : an application to Switzerland
MACCULI, Iris
Abstract
The present thesis examines the extent of poverty and welfare in Switzerland from a multidimensional perspective, using data from the Swiss Household Panel 1999-2007, a period of major economic and social changes in Switzerland. The thesis is divided into four chapters. The first two chapters examine the levels of deprivation and social exclusion, and their association to income poverty. The third chapter looks at the potential effects of income and multidimensional poverty on health status, with a focus on mental health. In this case, an instrumental variable method is used in order to take into account the endogeneity of explanatory variables. The last chapter provides a distributional analysis of earnings in Switzerland by examining the extent of inequality and polarization, and providing a decomposition analysis by socio-economic subgroups.
MACCULI, Iris. Essays on polarization, multidimensional poverty and health-related effects : an application to Switzerland. Thèse de doctorat : Univ. Genève, 2009, no. SES 711
URN : urn:nbn:ch:unige-142722
DOI : 10.13097/archive-ouverte/unige:14272
Available at:
http://archive-ouverte.unige.ch/unige:14272
Disclaimer: layout of this document may differ from the published version.
1 / 1
Essays on Polarization, Multidimensional Poverty and
Health-related Effects:
An Application to Switzerland
PhD Thesis
Iris MACCULI
Department of Economics University of Geneva
December 2009
2
Acknowledgements
First of all, I would like to thank my supervisor, Professor Yves Flückiger, for his support and guidance during all these years, and without whom this thesis would probably have never been completed. I am also particularly grateful to Professor Jacques Silber, for having accepted to be a jury member of this thesis. It is indeed an honour for me, given his great expertise in the areas of poverty measurement and decomposition analysis. I am similarly thankful to Professors Tobias Müller and Giovanni Ferro-Luzzi, for their collaboration and ongoing suggestions, as well as to Marcel Bilger, who worked with me in the last part of my thesis and made a valuable contribution to this work, bringing along his passion and competency. Many thanks also to Professor José Ramirez, for providing me with helpful comments and ideas, and to my friend and colleague Gabriela Flores, for the stimulating discussions we had during our academic path.
Similarly, I would like to thank all my other friends and colleagues, who have been there faithfully to encourage me. And last but not least, the most special gratitude goes to my family, for constantly giving me the motivation and support I needed, which were particularly important during most hard times. Finally, I would like to dedicate this thesis to my father, who certainly made me a richer person and who, I hope, would be proud of me today.
One could be well-off, without being well.
One could be well, without being able to lead the life he or she wanted.
One could have got the life he or she wanted, without being happy. One could be happy, without having much freedom.
One could have a good deal of freedom, without achieving much. We can go on.
[Amartya Sen, in Development as Freedom, 1999]
4
TABLE OF CONTENTS
INTRODUCTION ……….... 6
Chapter 1:
A multidimensional analysis of poverty and deprivation in Switzerland ……… 15
Chapter 2:
An assessment of social exclusion in Switzerland .………...46
Chapter 3:
Income, deprivation and mental health: What is the nature of the relation?... 72
Chapter 4:
Polarization and inequality of earnings in Switzerland. Outcomes and trends…102
CONCLUSION ……….. 134
Abstract
The present thesis examines the extent of poverty and welfare in Switzerland from a multidimensional perspective, using data from the Swiss Household Panel 1999- 2007, a period of major economic and social changes in Switzerland. The thesis is divided into four chapters. The first two chapters examine the levels of deprivation and social exclusion, and their association to income poverty. The third chapter looks at the potential effects of income and multidimensional poverty on health status, with a focus on mental health. In this case, an instrumental variable method is used in order to take into account the endogeneity of explanatory variables. The last chapter provides a distributional analysis of earnings in Switzerland by examining the extent of inequality and polarization, and providing a decomposition analysis by socio-economic subgroups.
6
Introduction
For decades, Switzerland has been a prosperous country, with one of the highest GDP per capita in the world and extremely low levels of unemployment.
In the 1990s, however, the country began to stagnate, a downturn from which the economy never fully recovered. The 1990s were indeed a decade of poor economic performance, unemployment rose rapidly, levelling out the rates of other small European countries. Slow growth continued throughout the beginning of this decade, accompanied by rapid social changes including demographic ageing, rising individualism, and changes in the household structure.
Important transformations also occurred on the labour market side. If on one hand the high flexibility of the Swiss market with the erosion of full-time positions made employment more uncertain, on the other hand an increasingly educated and skilled working population forced some people to leave the market.
As a result, a growing number of people began to rely on social assistance (3.3 percent in 2006, according to the Swiss Federal Statistical Office, that is three times the figures in 1988), putting the welfare system under increasing pressures.
Thus, a new sense of insecurity has arisen in Switzerland during the last decade, with increasing concerns about unemployment and poverty, conditions that are probably more difficult to accept in a traditionally affluent country like Switzerland. This thesis is devoted to enrich the existing literature by analysing specific aspects of individuals’ welfare during this period of social and economic turbulence.
The goal of the first chapter is to examine and explain poverty in Switzerland from a multidimensional perspective, between the end of the 1990s and today. In particular, we will try to understand the role played by income in explaining poverty. But also, what are the other relevant dimensions and how are they related to income? What are the most important socio-demographic factors driving individuals into poverty? Chapter 1 seeks to provide answers to these questions using data from the Swiss Household Panel.
It is widely accepted today that poverty is a multidimensional phenomenon, and that traditional income-based indicators are no longer sufficient to full capture individual living standards. Material and social aspects of life,
including housing conditions, possession of assets or participation in social activities are an essential part of individual well-being, and must be taken into account for a full understanding and a correct evaluation of poverty.
Since the pioneering work of Townsend (1979), a considerable amount of research has focused on multidimensional poverty measurement, and many empirical studies reported a substantial mismatch between poverty measured in terms of income and poverty measured using a variety of indicators, advocating the need for a multidimensional approach to poverty (Nolan and Whelan, 1996, Halleröd, 1996, Whelan et al. 2003). In Switzerland, however, studies on poverty are still scarce, Suter and Iglesias (2005) and Ferro-Luzzi et al. (2005) and Miceli (2006) representing the major contributions in this respect. This chapter aims at extending this literature by performing a comparative analysis of income poverty and multidimensional deprivation. For this purpose, we build on measures of deprivation, using a list of non-monetary indicators clustered into six basic dimensions of life. Deprivation indices for each dimension are obtained adopting the prevalence weighting procedure proposed by Desai and Shah (1988), where the weights reflect the proportion of individuals possessing the items in the population. We then examine levels and trends of multidimensional deprivation across dimensions, and test the consistency of our results with traditional income poverty measures. In the second part, we attempt to identify which factors lie behind poverty and deprivation in Switzerland. For this purpose, we use for the first time in this context a bivariate probit model, in order to quantify simultaneously the marginal effect of individual and household characteristics on poverty and deprivation by taking into account the positive correlation between these two outcomes.
The second chapter develops the analysis of the previous one by moving the emphasis on social exclusion. This has been the subject of considerable attention in recent years, especially in European countries where commitment was made to promote social cohesion and social inclusion in the Lisbon Treaty in 2002.
Despite the many attempts to find a consistent definition of social exclusion and the actual lack of agreement among academics, it is generally accepted that social
8
(2002) and Bossert et al. (2007), for instance, define social exclusion as the cumulative process of deprivation that is likely to affect participation of individuals in the society. While different interpretations are indeed plausible pointing to specific aspects of exclusion, more generally it seems reasonable to think that social exclusion may arise when a condition of deprivation persists or worsens over time. Building on this construct, we extend the analysis of the first chapter to social exclusion by examining levels of individual and social exclusion over time and across dimensions. In the second part, we perform a multivariate analysis in order to identify the determinants of social exclusion. Given the large number of null values characterizing the exclusion index, we propose a two-part model estimation allowing to assess the relative incidence of a series of individuals and household characteristics on both the probability and the levels of social exclusion, assuming that these might be driven by different factors.
In the third chapter we look at the potential effects of poverty and deprivation on mental health. There is wide recognition today that symptoms of anxiety and depression are associated with conditions of poverty and feelings of social exclusion. In a wealthy country like Switzerland, where social pressures and individual ambitions are widening, low socio-economic conditions appear to be important risk factors for the onset of depression and other common mental disorders. While this association between low socio-economic status and mental health is well documented, the nature of the relationship is not yet clearly understood. In the social sciences and epidemiological literature, for instance, there is ongoing debate on the direction of the causality (Dohrenwend et al., 1992, Johnson et al. 1999, Ritcher et al. 2001, Hudson, 2005, among others). While some argue that poverty, with its limited access to financial resources and access to health care, is in itself a stressful situation leading to mental illness (the social causation theory), others advocate the reverse, that is individuals genetically- prone to mental illness eventually “drift” down into poverty (the social selection or social drift theory). While some evidence has shown that the latter is more often observed in case of highly psychotic disorders like schizophrenia (Duhnam, 1965, Dohrenwend et al. 1992), the majority of studies support the hypothesis of social causation in the poverty and health relationship. The goal of our third chapter is then to assess the effect of income and multidimensional deprivation on health in Switzerland, with a particular focus on mental health.
A number of studies have reported a deterioration in psychological health in Switzerland in most recent years. Jägger et al. (2008), for instance, show that 25 percent of the population suffers today from some forms of depression, particularly among women. Though the government has recently committed itself to take action against depression and other mental diseases, empirical research in Switzerland is still lacking. Two studies are worth mentioning. Domenighetti et al.
(2000), who investigate the effect of job insecurity on health, and Vetter et al.
(2006), who examine the impact of financial deprivation on psychological well- being. Both these studies, however, perform logistic regressions analyses, which do not allow controlling for endogeneity in the model.
The original feature of this chapter is to use an instrumental variable method in order to account for the endogeneity of income and deprivation.
Moreover, the use of panel data allows us to control for unobserved heterogeneity, an issue that is particularly important in health outcomes.
The last chapter shifts the focus on distributional analysis and examines more specifically the extent of inequality and polarization of earnings distributions in Switzerland. As mentioned at the beginning of the introduction, labour markets went through a series of structural changes in recent years. After a period of increasing labour force that characterized the 1980s, the six-year recession of the early 1990s led to an unprecedented high unemployment rate, with a trend that continued throughout this decade (see Figure 1 below). Further, an increasing number of often low-paid part-time and temporary jobs made employment more uncertain. Immigration also increased in recent years, foreign residents making up about 25 percent of the total working population today (they were only 14 percent in 1980), and the majority of immigrants are today highly educated workers and their family members. Moreover, the Bilateral Agreements signed between Switzerland and EU members on the free movement of people and implemented in 2002 produced also important changes in the labour market structure.
10
Figure 1: Switzerland Unemployment Rate (in %)
Source: State Secretariat for Economic (Swiss Labour Force Survey)
Another important feature of the Swiss labour market is its segmentation as regards to gender, women working mainly part-time while men occupy more often full-time positions, resulting in increased earnings inequalities between men and women.
But then, how large are these inequalities? What is the impact of all those recent changes on earnings distributional patterns? Have distributions become more unequal and/or more polarized? How the between and within-group differences among subgroups contributed to explain the inequality and polarization of earnings? These are some of the questions that form the backdrop of our last chapter.
In recent years, polarization has received much attention, both theoretically and empirically, as an alternative measure to inequality in the analysis of income distributions. In particular, polarization reflects the clustering around local poles on the distribution that inequality measures fail to detect. In a more sociological perspective, it captures the tendency of economic agents to cluster into groups that are internally homogenous but increasingly different (or distant) from each other. In this construct, polarization may be potentially more relevant than inequality in explaining social conflicts and tensions within members of the society (Esteban and Ray, 1999).
While there is evidence on the extent of income inequalities in Switzerland (Leu et al. 2000, Gerfin, 1994, Piketty et al. 2007, Küng and Blank, 2000, among
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Total Men Women
others), all documenting an increase in inequality since the early 1980s, we are not aware of studies assessing the polarization of incomes. This paper brings a new contribution to this literature, by examining patterns of polarization and inequality between the end of the 1990s and today, providing a decomposition analysis by socio-economic subgroups.
12 References
ATKINSON, A.B., T. B. Cantillon, B. Marliner and B. Nolan (2002), Social Indicators. The EU and social inclusion, Oxford University Press, Oxford
BOSSERT, W., C. D'Ambrosio and V. Peragine (2007), "Deprivation and Social Exclusion", Economica, 74(296), 777-803
DESAI, M. and A. Shah (1988), "An Econometric Approach to the Measurement of Poverty", Oxford Economic Papers, New Series, 40(3), 505-522
DOHRENWEND, B.P. et al. (1992), “Socioconomic status and psychatric disorders: the causation-selection issue”, Science, 255: 946-52
DOMENIGHETTI, G., B. D’Avanzo and B. Bisig (2000), “Health effects of job insecurity among employees in Swiss general population”, Journal of Health Services, 30: 477-490
DUNHAM, H.W. (1965), Community and Schizophrenia: An Epidemiological Analysis. Wayne State University Press, Detroit
ESTEBAN, J. and D. Ray (1994), “On the Measurement of Polarization”, Econometrica, 62, 819-852
FERRO-LUZZI, G. Y. Flückiger and S. Weber (2005), “Multidimensional poverty: factor and cluster analysis”, chapter 4 in Quantitative Analysis and Multidimensional Poverty, edited by N. Kakwani and J. Silber, Palgrave Macmillan
GERFIN, M. (1994), “Income distribution, income inequality and life-cycle effects: A non-parametric analysis for Switzerland”, Swiss Journal of Economics and Statistics, vol. 130 (3): 509-522
HALLERÖD, B. (1996), "The truly poor: direct and indirect measurement of consensual poverty in Sweden", Journal of European Social Policy, 5(2), 111-129
HUDSON, C.G. (2005), "Socioeconomic status and mental illness: Tests of the social causation and selection hypotheses", American Journal of Orthopsychiatry, 75(1): 3-18
JONHSON, J.G., P. Cohen, B.P. Dohrenwend et al. (1999), “A longitudinal investigation of social causation and social selection processes involved in the association between socioeconomic status and psychatric disorders”, Journal of Abnormal Psychology, 108: 490-499
JÄGER, M. P. Sobocki and W. Rössler (2008), “Cost of disorders of the brain in Switzerland. With a focus on mental disorders”, Swiss Medical Weekly, 138: 4-11 KÜNG GUGLER, A. and S. Blank (2000), “Inégalité des salaires en Suisse: pas d’augmentation sensible durant les années 90”, Swiss Journal of Economics and Statistics, Vol. 136(3): 307-317
LEU, R.E., C. Ernst and M. Gerfin (2000), “Inequality Trends in the Swiss Income Distribution”, Swiss Journal of Economics and Statistics, Vol. 136(3):
289-305
MICELI, D. (2006), “Multidimensional and Fuzzy Poverty in Switzerland”, chapter 3, in Fuzzy Set Approach to Multidimensional Poverty Measurement, edited by A. Lemmi and G. Betti, Springer US
NOLAN, B. and C.T. Whelan (1996), Resources, Deprivation and Poverty, Oxford University Press, Clarendon Press, Oxford and New York
PIKETTY, T., F. Dell and E. Saez (2007), “Income and wealth concentration in Switzerland over the twentieth century”, chapter 11, in “Top incomes over the twentieth century: a contrast between continental European and English-speaking countries”, Oxford: Oxford University Press, 472-500
14
social selection and social causation models”, British Journal of Psychiatry, 178:
s84-s90
SUTER, C. and K. Iglesias, (2005), “Deprivation and Well-being: Switzerland in a comparative perspective”, in Hanspeter Kriesi, Peter Farago, Martin Kohli and Milad Zarin (eds.), Contemporary Switzerland: The End of Exceptionalism
TOWNSEND, P. (1979), Poverty in the United Kingdom, Harmondsworth, Penguin
VETTER, S. J. Endrass, I. Schweizer, H. Teng, W. Rossler and W.T. Gallo (2006), “The effect of economic deprivation on psychological well-being of working population in Switzerland”, BMC Public Health, 6: 223
WHELAN, C.T. et al. (2003), "Persistent income poverty and deprivation in the EU: an analysis of the first three waves of the ECHP", Journal of Social Policy, 32(1), 1-18
Chapter 1:
A multidimensional analysis
of poverty and deprivation in Switzerland
Abstract
In this paper we provide empirical evidence on multidimensional aspects of poverty in Switzerland, using data from the Swiss Household Panel for the period 1999-2004. We define measures of multiple deprivation using a range of non-monetary indicators and test the consistency of our results with traditional income poverty measures. We then perform a bivariate probit analysis to determine the simultaneous incidence of a set of observed characteristics on the probability of both poverty and deprivation. Results show that deprivation decreased over the period, with some notable differences among dimensions, people appearing more highly deprived in the financial and housing dimensions.
The low degree of overlap between income and other dimensions of deprivation underlines the importance of the measure used and its implications for social policy targeting.
16 1. Introduction
A few years ago, Swiss Minister Pascal Couchepin warned that the Swiss society could be sitting on a bomb if it doesn’t do more to help people getting out of poverty. He added that even if the great majority of the population was doing alright, the authorities could not ignore the conditions faced by part of the population (Neue Zürcher Zeitung, November 2006).
For several decades, Switzerland has been one of the countries with the highest GDP per capita, with unemployment and inflation rates well below those of other developed countries. In recent years, however, the economic and social context has changed. The 1990s were a decade of poor economic performance.
Unemployment rose rapidly and an increasing number of people began to rely on social assistance (3.3 percent of the population in 2006, according to the Federal Statistical Office). The beginning of the twenty-first century was also characterized by slow economic growth and increasing tensions in the labour market, fostering the debates on poverty and welfare issues among policy makers and social institutions1.
Poverty is indeed a multidimensional phenomenon. If income is an important dimension as it provides the material resources to fulfil a subsistent life, individual well-being depends also on other attributes, including for instance housing conditions, education, health or participation in social activities. As Amartya Sen has put it, “the role of income and wealth has to be integrated into a broader and fuller picture of success and deprivation” (Sen, 1981).
While multidimensional poverty has been subject to much investigation in the international literature, empirical evidence in Switzerland is still scarcer, Suter and Iglesias (2005), Ferro-Luzzi et al. (2005) and Miceli (2006) representing the major contributions in this respect. While the latter adopted the fuzzy set approach to examine multidimensional poverty using data from the Swiss Household Panel, Ferro-Luzzi et al. (2005) carried out a factor and cluster analysis on the same database to identify the relevant dimensions and the groups most affected by poverty. They identified four main dimensions of poverty in Switzerland, namely financial poverty, poor health, bad neighbourhood and social exclusion. Finally, in a comparative study, Suter and Iglesias (2005) examined material and social
1 Kehoe and Ruhl (2005) argue that recession started well before the 1990s, showing that indeed since the mid-1970s Switzerland is suffering an authentic period of “great depression”, in terms of decreasing growth of productivity and GDP per capita.
deprivation using the Euromodule data for the year 2000. Their results showed that Switzerland enjoys in fact the highest living standards and subjective well- being compared to other Western countries, including Germany and Spain. They also found that the Swiss are not much demanding in terms of possession of durables while they give larger importance to savings. Finally, long-term unemployed, divorced and disabled persons were reported to be the groups with the highest risk of deprivation in Switzerland.
This paper attempts to extend this line of research by providing a comprehensive analysis of income and multidimensional poverty in a comparative perspective, for the period 1999-2004. For this purpose, we build on measures of material and social deprivation and look at their consistency with traditional income poverty measures. In the second part, we try to identify the major determinants of income poverty and deprivation simultaneously, while examining the link between their risk factors. For this purpose, we make use of a bivariate probit model, which takes explicitly into account the correlation between the two outcomes. To our knowledge, this is done for the first time in this context, as most of the previous studies have generally performed OLS regressions (Layte et al.
2001a), or logistic estimations (Layte et al. 2001b, Whelan et al. 2003).
Assuming that a correlation exists between income and deprivation, we suspect however to find a low degree of consistency between these two outcomes, meaning that individuals who do fall below the income poverty line are not necessarily poor in terms of living conditions. Current income is only a transitory flow of cash, and people might well have been cumulating assets or good housing conditions as to ensure them a decent life. Or it might be that individuals are actually poor in terms of income but do not “feel” that they are poor, in terms of subjective well-being (Nolan and Whelan, 1996, Layte et al. 2001). Such analysis would help us identifying who exactly are the poor and in what dimensions individuals are poor, a result that should have important policy implications in terms of poverty targeting.
The rest of the paper is structured as follows. The next section provides a short review of the literature on multidimensional poverty. Section 3 discusses the definitions and methodologies used in the paper, and presents the estimation
18 2. Relevant literature
Multidimensional poverty analysis has received a considerable amount of attention in recent years, both theoretically and empirically. Implicit in this approach is the idea that income alone is not a sufficient nor an appropriate measure of welfare, being often underestimated and of irregular nature, and that composite indicators should be used to take account of the complexity and multidimensional nature of poverty, including housing and environmental conditions, material deprivation and lack of social relations. In the words of Atkinson, “there is widespread agreement that deprivation is multidimensioned. It is not enough to look only at income poverty; we have to look at other attributes.”
(Atkinson et al. 2002). The multidimensional nature of poverty seems nowadays to be well-established and fully integrated into the social and political discourse of major institutions2, the best example being probably the United Nations Millennium Development Goals project to end poverty by 2015.
In 1979, Townsend was the first to define poverty in terms of relative deprivation, opening the way to a multidimensional approach to poverty. Inspired by the work of Runciman (1966)3, he defined relative deprivation as "the absence of those diets, amenities, standards, services and activities, which are common or customary in society. People are deprived of the conditions of life which ordinarily define membership of society, if they lack or are denied resources to obtain access to these conditions of life and so fulfil memberships of society, they are in poverty.” (Townsend 1979:915). In the following years, Ringen (1988)
2 The European Commission declared that “(…) the poor shall be taken to mean persons, families and groups of persons whose resources (material, cultural and social) are so limited as to exclude them from the minimum acceptable way of life in the member state in which they live” (EEC, 1985). Similarly, the United Nations, during the World Summit for Social Development, defined poverty in the following terms: “The lack of income and productive resources to ensure sustainable livelihoods: hunger and malnutrition; ill health; limited or lack of access to education and other basic services; increased morbidity and mortality from illness;
homelessness and inadequate housing; unsafe environments and social discrimination and exclusion. It is also characterised by lack of participation in decision-making and in civil, social and cultural life.” (The Copenaghen Declaration, UN, 1995, p.57)
3 Runciman (1966) developed the idea that an individual’s feeling of deprivation arises when comparing her situation with that of others who are better off: « The magnitude of a relative deprivation is the extent of the difference between the desired situation and that of the person desiring it » (Runciman, 1966).
acknowledged that poverty must be intended as “a state of general deprivation characterised by both a low standard of consumption and a low level of income”.
In 1985, Sen introduced a new perspective on multidimensional poverty analysis, defining the individual well-being as “an index of the person’s functionings”, where functionings represent all the things a person can manage to do or be, and capabilities are the freedom of a person to choose her functionings (Sen, 1985). Sen’s capability approach inspired much of the literature on multidimensional poverty that followed, including the construction of the Human Development Index by the UNDP, which aggregates for each country the functioning achievements in terms of life expectancy, real GDP per capita and educational levels (UNDP, 1990).
European countries have also developed a long tradition on multidimensional poverty measurement, principally based on the use of material and social indicators as a more direct measure or living standards. This approach, inspired by Townsend’s work, has been applied in a large number of empirical studies, all advocating the use of non-monetary indicators in the measurement of poverty as they are also more stable over time (Mack & Lansley 1983, Halleröd, 1996, Nolan & Whelan 1996, Muffels 1993, Bradshaw and Finch, 2003, Whelan et al., 2003 among others). The idea is to aggregate a set of indicators into a number of relevant dimensions, and count the number of dimensions in which individuals are deprived. Many of these papers focused on the relation between income – an indirect measure of well-being - and multidimensional deprivation – a more direct measure of well-being, finding that such relationship was far from perfect. They showed, for instance, that while some low-incomes may not be suffering poor living conditions, other individuals who are above the poverty line are probably experiencing deprivation in other aspects of life.
Along with this empirical evidence, the axiomatic approaches to multidimensional poverty began to gain ground in the last decade. Brandolini and D’Alessio (1998) applied Sen’s capability approach to the multidimensional analysis of deprivation and inequality, while Tsui (2002) generalized the class of subgroup consistent poverty indices introduced by Foster and Shorrocks (1991) to the multidimensional context. Atkinson (2003) and Bourguignon and
20
multidimensional poverty making an empirical comparison of various approaches, while Kakwani and Silber (2008a, 2008b) published a collection of articles on multidimensional poverty measurement covering also different approaches, including information theory, fuzzy sets, as well as the axiomatic and the social welfare theories.
3. Methodological issues
3.1. Concepts and definitions
In this essay, we use two different poverty measures. The first is a unidimensional measure, based on a level of income - usually 50 or 60 percent of the mean or median income – that is considered the threshold below which individuals are considered to be poor. It is the traditional and most commonly used relative income measure, applied in many poverty studies. Despite its limitations related to the arbitrary definition of a poverty line and to its reliance on the income distribution, this measure is straightforward and allows for comparison of poverty levels across countries. In our analysis, the variable of interest is the annual household income, net of taxes and social contributions, and adjusted using the OECD-modified equivalence scale in order to correct for households’
different size and needs. We set the poverty line at 50 percent median income and use the individual as the unit of analysis, although many variables are taken at the household level.
The second measure is multidimensional and based on a range of non- monetary indicators covering various social and material dimensions of welfare.
The ability of households to afford such items is at the base of this approach (Townsend, 1979). Hence poverty, defined here in terms of multiple deprivation, arises from the “enforced” absence of items and activities4, which are considered to be customary in the society in which we live.
Under this construct, deprivation appears to be a more sensitive measure of
“command over resources” than current income, as it provides a more accurate assessment of individuals’ consumption patterns and living conditions (Whelan et
4 « Enforced » in terms of the lack of financial resources.
al. 2003). The next section describes in details the construction of the deprivation measure.
3.2. Constructing indices of deprivation
If using income to identify the poor does not involve particular difficulties, except for the choice of the poverty line, measuring poverty in the multidimensional space requires a certain number of decisions as to how to select the relevant indicators and dimensions of poverty, as well as the weights to apply in the aggregation process. All or at least most of these decisions, though inevitably arbitrary, rely on the data availability and on the objective of the study.
The first step in the analysis relies on the choice of indicators. Given the information available in our dataset and following the suggestions given by Atkinson et al. (2002), we select a set of 27 items and activities reported in Table 1. Although not exhaustive, this list of indicators is selected on the basis of what we consider to be common items in Switzerland today and whose lack would be considered as a socially-perceived deprivation. The "enforced lack" criterion introduced for the first time by Mack and Lansley (1985) is adopted whenever this was possible, meaning that only individuals who could not afford the item because of lack of resources would be considered as deprived5. However, as the perception of necessity is intrinsically influenced by subjective tastes and attitudes6 7, and given the arbitrariness embedded in the choice of indicators, as mentioned above, we test the reliability of the indicators using the index of Cronbach's Alpha, a statistic measuring the internal consistency (or homogeneity) of the underlying indicators and expressed as a function of the number of items and the average of all covariances between them. Overall, the values obtained are relatively high (around 0.8 for all indicators), which is indicative of a reliable measure. The range of indicators is then aggregated into six dimensions, covering
5 Sometimes, as for the housing dimension, this information is not available and in this case the simple shortage is accounted as deprivation.
6 Mc Kay (2004) shows how some families who cannot afford necessities often possess a number of non-necessities (generally luxury goods), and these atypical preferences reveal that « such families may be classified as poor using deprivation indicators, when it might be more accurate to
22
the following areas of deprivation: basic deprivation, housing conditions, possession of durables, lifestyle, financial and subjective deprivation. The choice of dimensions is also central in multidimensional poverty analysis. What dimensions should be more relevant? The answer is indeed not easy. Our choice is driven by main suggestions given in the existing literature (Atkinson et al., 2002 and Layte et al. 2001) and by what we believe to be important dimensions of life in the Swiss contemporary society8.
A further step in multidimensional poverty measurement concerns the choice of the weighting structure. In the process of aggregation, weights should reflect the importance of each indicator in the overall deprivation measure. Rather than using equal weights, whereby all items are assigned the same value, we apply a method where the weights reflect the spread of each item among the population9. For this purpose, we follow the prevalence weighting procedure introduced by Desai and Shah (1988), whereby each weight corresponds to the share of individuals possessing the corresponding item among the entire population. This “frequency-based” method has the advantage of giving more weight to situations of minor deprivation among the population. To be more specific, if an individual does not possess an item that is owned by the majority of the population (say, a television) because of a lack of resources, such a deprivation will be given a higher weight compared to the case where the same individual does not possess an item that is not very common in the population (say, a garden)10.
After aggregation, the partial index of deprivation for each dimension m (m=1,2, …, M) is therefore given by the following expression:
100
∑
*∑
=
j m j j
m ji m j m
i w
I w
D [1]
8 We explicitly exclude education from the analysis, as one of our goals is in fact to assess the effect of education on deprivation.
9 In their paper, Suter and Iglesias (2005) used the proportional deprivation index introduced by Halleröd (1985), in which the weights reflect the proportion of individuals who consider a specific item as absolutely necessary.
10 As a robustness check, it would certainly be of interest to run some sensitivity analysis by using more than one approach, and see how, for instance, the results differ when using equal weighting.
where Imji is a dichotomous variable taking value of 1 if individual i lacks item j (j=1,2,…,Jm) of dimension m, and 0 otherwise, and is the proportion of people not lacking item j of dimension m11. We then normalize to obtain a partial index of deprivation (one for each dimension), with values varying between 0 and 100.
These values, or scores, can therefore be interpreted as the percentage of deprivation experienced by each individual, compared to a maximum value of 100 that would be given to an individual experiencing deprivation in all items in that specific dimension. A total index of deprivation is then obtained by weighted sum of all partial indices, using the same aggregating structure as in [1], but with weights being in this case an average of the weights applied in partial indices12. Hence, for each individual, the total deprivation index is given by:
∑
∑
=
= =M m
m m i M
m m
Ti
w D w D
1
1 [2]
where Dim represent the index of deprivation in each dimension m, and w m the corresponding average weights.
3.3. The empirical model
The goal in this section is to identify the factors affecting the probability of poverty and deprivation in Switzerland, and trying to understand if a link exists between their risk factors. For this purpose, we build a model with two binary dependent variables:
- income poverty: Y1, equal to 1 if the individual is below the poverty line set at 50 percent median income, and 0 otherwise
- multiple deprivation: Y2, equal to 1 if the individual is below the deprivation line, and 0 otherwise.
11 Applying the enforced lack criterion, it is important to bear in mind that this group includes all
m
wj
24
In order to allow comparison throughout the analysis, the deprivation line is set at a level giving the same proportion of deprived as we have of poors13.
Assuming correlated errors and normality in the distribution functions, we use a maximum-likelihood bivariate probit model to estimate the relative incidence of the independent variables on each of the two outcomes. The independent variables include a range of characteristics specific to the individual (age, gender, sex, citizenship), market-related (occupation, education) and life- related (household type, civil status, community typology, language region).
4. Data
Our analysis is based on data drawn from the first six waves of the Swiss Household Panel (SHP), covering the period 1999 through 2004. The SHP started in 1999 as part of the project Living in Switzerland run by the National Science Foundation, the Swiss Federal Statistical Office and the University of Neuchâtel.
The original sample consists of around 13’000 respondents aged 15 and over and belonging to more than 5’000 households. The annual questionnaires, based on a Computer-Assisted Telephone Interview, are held in French, German or Italian depending on the preference of the respondent. People are interviewed on their individual and household characteristics in a broad range of economic and social subjects, from all sources of income to various living aspects, including housing, health and leisure activities. Subjective assessment of well-being is given along with factual information14.
Cross-sectional analysis is based on data from 1999, on a total sample of 12’931 individuals. We use the individual as the unit of analysis, although most of the information is collected at the household level. Finally, taking into account the initial non-response, we use the appropriate cross-sectional sample weights in order to correct for sample selection bias.
13 While the choice of any threshold in welfare analysis is inevitably arbitrary, we select a deprivation line allowing consistent comparisons with levels of income poverty (Layte et al., 2001b, Whelan et al. 2003).
14 Detailed information on the panel structure is given in the Appendix.
5. Empirical results
5.1. Income and multiple deprivation
Table 2 presents the distribution of population by levels of deprivation in each dimension at the beginning of the period (1999). As we can see immediately from the table, differences across dimensions are significant. While more than 90 percent report no deprivation (score=0) in the possession of durables, about two- third of the population suffer deprivation in the housing dimension, with more than 9 percent reporting high levels of deprivation in this area (score over 40).
From a policy perspective, such differences are indeed interesting, suggesting that each dimension is a phenomenon per se reflecting in fact different facets of deprivation.
Looking at the relation between income and multiple deprivation in Table 3, we find that correlations are very low, consistent with results found in previous studies15, and only two dimensions, namely the financial and lifestyle dimensions, show slightly higher correlations, a result somehow expected given the nature of the indicators included in these dimensions. Housing appears the least correlated with income, in line with findings reported by Ayllòn et al. (2004) for Spain. The link between income and deprivation is further illustrated in Figure 1, which plots the distribution of deprivation by income decile. As we can see, the gradient for the basic and the financial dimensions is much steeper but also less smooth than the other dimensions, while housing deprivation appears once again the least correlated with current income. These findings seem to confirm some predictions from economic theory, since housing is more often associated with decisions of investment and savings, and therefore related possibly to a form of long-term income rather than current income. Exploring a little further the degree of overlap between income and the other dimensions of deprivation, we calculate the mean deprivation scores around points on the income distribution usually taken as poverty thresholds (Table 4). The outcomes are rather surprising, as some individuals appear to be more deprived when their income is between 40 and 50 percent of the median, than when this is below 40 percent of the median. This result is a further proof that people who are identified as poor in terms of income
26
are not necessarily poor in other dimensions of life. Also, it acknowledges that the identification of the poor depends dramatically on the measure adopted.
Finally, correlations between dimensions are in all points positive but relatively low, most likely due to the different nature of the indicators included in each dimension, and suggesting that each dimension does reflect different aspects of deprivation (results are reported in the Appendix).
5.2. Trends in poverty and deprivation
In this section we examine the evolution of income poverty and deprivation between 1999 and 2004. Table 5 reports the mean deprivation scores for each dimension of deprivation with estimates for income poverty expressed by the head count ratio. Although the levels for these measures are not comparable, as they differ in nature, we can compare how they evolve over time. We can see that conditions improve particularly in the housing dimension, the deprivation score decreasing from 16 to around 11 over the period, with a striking fall at the beginning of the period, while it remains more or less stable for the other dimensions. The lowest level of deprivation is observed in the possession of durables, with a total mean score below 3. Income poverty follows a different pattern. The head-count ratio decreases until 2001, possibly as an effect of the first recovery of the late 1990s, and then moves up again by the end of the period.
Interestingly, a similar trend is only observed in the financial area, the same showing the highest correlation with income and the highest levels of deprivation at the beginning of the period. Finally, changes in total deprivation are observed particularly in the lower tail of the distribution, the percentage of non-deprived having more than doubled in six years (Figure 2).
5.3. Results from multivariate analysis
As a first step to regression analysis, Table 7 reports the different profiles of poverty and deprivation, according to the threshold defined in the previous section. In 1999, 87 percent of the population is neither poor nor deprived, and this proportion increases over the period. At the other extreme, 1.24 percent of individuals was simultaneously poor and deprived, a condition often regarded as
“consistent” or “persistent” poverty (Layte et al. 2001b, Whelan et al. 2003).
Table 8 presents the estimates from the bivariate probit regressions performed for 1999. In particular, we report the marginal effects of a set of explanatory variables on the four profiles of poverty described in Table 7, all other factors held constant. This econometric strategy allows to directly comparing the relative incidence of each explanatory factor on the relevant outcome. Employment and education appear to be the highest predictors for both poverty and deprivation. In particular, unemployment increases by almost 16 percent the probability of being deprived but not poor, while education (secondary or tertiary compared to primary) allows individuals to escape poverty and deprivation by around 13 percent. Employment plays indeed a major role in protecting against adverse outcomes. An individual who is unemployed, while being excluded from work, is also denied access to resources, assets and participation in many other aspects of life. Quite surprisingly however, being retired or working part-time does not have a significant effect on neither of the two outcomes.
Being foreign does not seem to affect the risk of poverty, but has a significant positive effect on multiple deprivation, in line with results found by Ferro-Luzzi et al. (2005). The household structure plays also a role on both outcomes. In particular, large families (three or more children) and single-parent families are much more likely to experience poverty, though their effects remain small in magnitude (around 3-4 percent). Finally, being married, compared to a single, increases by 18 percent the likelihood of escaping both poverty and deprivation, suggesting that not only the emotional support of a partner, but her material support would tend to be an important protective factor against adversity.
6. Concluding remarks
Poverty is becoming increasingly relevant in a country like Switzerland traditionally regarded as a “rich” country. In this paper we examine the extent of poverty from a multidimensional perspective, drawing data from the Swiss Household Panel for the period 1999-2004. In particular, we compare and test
28
terms of low-income, against 7.4 percent who are deprived in several aspects of living. In comparison, consistent poverty – the simultaneous condition of poverty and deprivation – is however relatively small (1.24 percent). Among the various dimensions, the Swiss seem to be better off in the possession of durables, whilst the highest levels of deprivation are found in housing, where more than 65 percent of individuals experience some degree of deprivation in this area. Housing is also the least correlated with income, in line with results found in previous studies, and consistent with economic theory suggesting that housing is often the result of investment choices and, therefore, more related to savings and cumulated wealth rather than current income. The low degree of overlap between income and the other dimensions of deprivation confirms also the importance of analysing each of these dimensions separately as they all reflect different aspects of individual welfare. Finally, we identify in employment and education the major predictors of both poverty and deprivation, a result suggesting that policies should probably focus on improving access to labour markets and investments in human capital.
Direct social transfers to the poor are important, but they will probably not solve all the evils. Housing conditions are also an important aspect of well-being, and more than half of the population suffer some form of deprivation in this area.
Urbanization policies and easing access to credit would probably contribute to improve environmental and housing conditions for many.
From a social policy perspective, these results should therefore help policy makers to elaborate policies targeted to specific dimensions of deprivation and specific groups of the population. Indeed, more work needs to be done. One further step would be to extend the analysis to other dimensions, including for instance health and education. It would be also interesting to incorporate a dynamic measure of deprivation, which would help identifying possible sequences of deprivation while contributing to understand its real causes.
References
ANAND, S., and A.K. Sen (1997), Concepts of Human Development and Poverty: A Multidimensional Perspective, UNDP, New York
ATKINSON, A.B., T. B. Cantillon, B. Marliner and B. Nolan (2002), Social Indicators. The EU and social inclusion, Oxford University Press, Oxford
ATKINSON, A.B., (2003), “Multidimensional deprivation. Contrasting social welfare and counting approaches”, Journal of Economic Inequality, 1, 51-65 AYLLÒN, S. et al. (2004), “Caracterizaciòn de la privaciòn y de la pobreza en Cataluña”, Document de Treball, 04.10, Universitat Autònoma de Barcelona, Spain
BOURGUIGNON, F., and S.R. CHAKRAWARTY (2003), “The measurement of multidimensional poverty”, Journal of Economic Inequality, 1(1), 25-49
BRADSHAW, J., and N. Finch (2003), “Overlaps in dimensions of poverty”, Journal of Social Policy, 32(4), 513-25
BRANDOLINI, A., and G. D’Alessio (1998), “Measuring well-being in the Functioning Space”, mimeo, Banca d’Italia Research Department, Rome
DESAI, M. and A. Shah (1988), "An Econometric Approach to the Measurement of Poverty", Oxford Economic Papers, New Series, 40(3), 505-522
DEUTSCH, J. and J. Silber (2005), “Measuring multidimensional poverty: An empirical comparison of various approaches”, Review of Income and Wealth, 51(1), 145-174
EEC, “On Specific Community Action to Combat Poverty” (Council Decision of
30
FERRO-LUZZI, G. Y. Flückiger and S. Weber (2005), “Multidimensional poverty: factor and cluster analysis”, chapter 4 in Quantitative Analysis and Multidimensional Poverty, edited by N. Kakwani and J. Silber, Palgrave Macmillan
FOSTER, J. and A. Shorrocks (1991), “Subgroup consistent poverty indices”, Econometrica, 59: 687-709
GREENE, W. (2007), Econometric Analysis, 6th Edition, Prentice Hall, New York HALLERÖD, B. (1996), "The truly poor: direct and indirect measurement of consensual poverty in Sweden", Journal of European Social Policy, 5(2), 111-129 KAKWANI, N., and J. Silber (2008a), The many dimensions of poverty, Palgrave Macmillian
KAKWANI, N., and J. Silber (2008b), Quantitative approaches to multidimensional poverty measurement, Palgrave Macmillan
KEHOE, T.J. and K.J. Ruhl (2005), “Is Switzerland in a Great Depression?”, Review of Economic Dynamics, 759-775
LAYTE, R. et al. (2001a), "Explaining levels of deprivation in the European Union", Acta Sociologica, 44, 105-121
LAYTE, R., et al. (2001b), “Persistent and consistent poverty in the 1994 and 1995 waves of the European Community Household Survey, Review of Income and Wealth, 47(4), 427-449
MACK, J. and S. Lansley (1983), Poor Britain, Allen Unwin, London
McKAY, S. (2004), “Poverty or preferences: what do ‘consensual deprivation indicators’ really measure?”, Fiscal Studies, 25(2), 201-223
MICELI, D. (2006), “Multidimensional and Fuzzy Poverty in Switzerland”, chapter 3, in Fuzzy Set Approach to Multidimensional Poverty Measurement, edited by A. Lemmi and G. Betti, Springer US
MUFFELS, R. (1993), "Deprivation standards and style of living indices", in Bergman and Cantillon (eds), The European Face of Social Security, Avebury, Aldershop, 43-59
NOLAN, B. and C.T. Whelan (1996), Resources, Deprivation and Poverty, Oxford University Press, Clarendon Press, Oxford and New York
PIACHAUD, D. (1981), “Peter Townsend and the Holy Grail”, New Society, (10), 419-431
PAYNE, S. and C. Pantazis (1997), “Poverty and gender”, in D.Gordon and C.
Pantazis (eds.) Breadline Britain in the 1990s, 97-110, Ashgate, Aldershot
RINGEN, S. (1988), Direct and Indirect Measures of Poverty", Journal of Social Policy, 17(3), 351-366
ROOM, G. (1995), Beyond the threshold: The measurement and analysis of social exclusion, The Policy Press, Bristol
RUNCIMAN, W.G. (1966), Relative deprivation and social justice, Routledge, London
SEN, A. (1981), Poverty and famine: an essay on entitlement and deprivation, Clarendon Press, Oxford
SEN, A. (1985), Commodities and Capabilities, North-Holland, Amsterdam SUTER, C. and K. Iglesias, (2005), “Relative deprivation and well-being:
32
TOWNSEND, P. (1979), Poverty in the United Kingdom, Harmondsworth, Penguin
TSUI, K. (2002), “Multidimensional poverty indices”, Social Choice and Welfare, 19: 69-93
U.N. (1995), The Copenaghen Declaration and Program of Action: World Summit for Social Development, 6-12 March 1995, New York, NY: United Nations Department of Publications
UNDP (1990), Human Development Report, Oxford University Press, New York, NY
WHELAN, C.T. et al. (2003), "Persistent income poverty and deprivation in the EU: An analysis of the first three waves of the ECHP", Journal of Social Policy, 32(1), 1-18
WHELAN, C.T., R. Layte and B. Maître (2005), “Understanding the mismatch between income poverty and deprivation: A dynamic comparative analysis”, European Sociological Review, 20(4), 287-302
Table 1 : Indicators used to construct deprivation indices (% of enforced lack), 1999
Items % Weight
Basic deprivation (B)
Arrears in payments in the last 12 months 11.94 0.88
Cannot afford to invite friends at least once a month 2.58 0.70 Cannot afford a meal at a restaurant at least once a month 14.65 0.54
Cannot afford to go to the dentist if needed 3.05 0.97
Durable goods (D)
Cannot afford a car 3.09 0.89
Cannot afford a colour TV 0.26 0.94
Cannot afford a private washing machine 1.04 0.69
Cannot afford a dishwasher 2.92 0.71
Cannot afford a computer 4.83 0.68
Housing conditions (H)
Accommodation in bad condition 31.59 0.68
Accommodation too small 15.15 0.85
Accommodation badly heated 8.82 0.91
Accommodation with noisy external environment 20.56 0.79
Accommodation: problems with pollution, traffic or industry 14.69 0.85 Accommodation with violence or vandalism around the house 13.70 0.86 Life-style and social deprivation (LS)
Cannot afford one-week holiday away from home per year 7.75 0.83
Cannot afford a home with a garden or terrace 2.24 0.83
Cannot afford a second home 29.58 0.16
Cannot afford internet access from home 6.62 0.33
Financial deprivation (F)
Unable to save 100CHF min per month 13.48 0.83
Unable to save for retirement saving schemes ("3rd pillar") 11.96 0.65 Payments of monthly premiums in the last 12 months 15.40 0.85 Financial situation unmanageable (unable to make ends meet) 23.24 0.77
Reception of financial help 19.67 0.80
Subjective deprivation (S)
Satisfaction with financial situation of household 18.16 0.93
Satisfaction with standard of living 9.88 0.90
Satisfaction with accommodation 6.77 0.82
Note: All variables are dichotomous, taking value of 1 if household does not have or cannot afford such items, 0 otherwise.
34
Table 2: Distribution of population by levels of individual deprivation, 1999 (%)
Segments
Dimensions
Total (T) Basic
(B)
Housing (H)
Durables (D)
Lifestyle (LS)
Financial (F)
Subjective (S)
0 75.93 34.44 90.77 64.84 51.90 75.60 14.68
1-20 8.94 38.45 4.93 23.09 10.03 - 68.83
21-40 8.65 18.19 2.73 5.55 19.52 15.66 13.49
41-60 5.13 6.54 1.47 4.58 12.29 - 2.67
61-80 0.99 2.12 0.06 1.53 5.10 7.28 0.33
81-100 0.35 0.26 0.03 0.41 1.16 1.46 -
Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 Mean
Std Dev
7.48 (15.50)
16.73 (16.66)
2.32 (8.08)
7.08 (14.73)
16.74 (21.92)
11.28 (22.25)
10.40 (11.31)
(Obs: 12931)
Note: For each dimension, the deprivation score is divided in quantiles representing levels of deprivation, from null (0) to maximum level of deprivation (100).
Table 3: Correlation between deprivation dimensions and household disposable income (1999)
(Obs: 10055). All coefficients are significant at 5 percent level.
Basic Durables Housing Lifestyle Financial Subjective Total
Income -0.20 -0.15 -0.04 -0.19 -0.26 -0.18 -0.25
Logincome -0.27 -0.22 -0.05 -0.25 -0.34 -0.24 -0.34
Figure 1: Distribution of deprivation, by income decile
Note: Mean deprivation scores are given for each dimension.
Distribution of deprivation, by income decile (1999)
Note: Mean deprivation scores are given for each dimension.
